A life history calendar allows mapping personal events to improve recall of past events in retrospective data collection. We estimated whether the use of a life history calendar was related to data completeness and accuracy. Participants in a case-control study on inflammatory bowel disease in Quebec, Canada in 2021, were invited to complete a preparatory life history calendar and encouraged to consult it during data collection. For data completeness, associations between life history calendar preparation/consultation frequency and number of missing values were estimated using negative binomial regression (Sample Mean Ratio, SMR) with inverse-probability-weighting to balance participants’ characteristics across life history calendar preparation/consultation groups. One hundred and thirty variables were considered. For accuracy, parents’ age at participants’ birth and number of older siblings, the only variables available from both the questionnaire and Birth Registry, were compared for agreement according to life history calendar preparation. Of 2727 participants, 48% prepared the life history calendar. During data collection, 27%, 48%, and 25% consulted it never, sometimes, often/always, respectively. The overall proportion of missing values was low (0.7%). Life history calendar preparation (vs. not) was associated with a 22% decrease in number of missing values (SMR = 0.78; 95% CI: 0.64–0.96). There were fewer missing values with greater consultation frequency, with 26% (SMR = 0.74; 95% CI: 0.56–0.98), 31% (SMR = 0.69; 95% CI: 0.55–0.86) and 59% (SMR = 0.41; 95% CI: 0.29–0.56) fewer missing values among those who never, sometimes, often/always consulted it, respectively, vs. not having completed a life history calendar. Life history calendar preparation was associated with slightly better agreement between self-reported and registry data for mother’s (5% higher) and father’s (4% higher) age at participant’s birth, but not number of older siblings (1% lower). Life history calendar preparation/consultation were associated with higher data completeness. However, the assessment of data accuracy was limited due to the small number of available variables.
Yu et al. (Thu,) studied this question.